RECURRENT SELECTION AND OVERDOMINANCE 469 



where k > \, thus providing the ])ositive mild regression of offspring on par- 

 ent, the heritability which so many have taken as strong evidence against 

 k> I. 



Many traits of the corn plant are mostly independent of genes concerned 

 with yield. Many others may be optimum for yield at intermediate points 

 genotypically as well phenotypically. It should hardly seem surprising if, 

 subsequent to intense selection for yield, we should find evidence of inter- 

 mediate gene frequencies and very little inbreeding depression or heterosis 

 with such characters. An intermediate optimum may place some premium on 

 aA, but hardly to the extent of explaining the evident heterosis of corn 

 yields, so far as I can see. 



Evidence cited here of overdominance in the genetics of grain yield of corn 

 consists of: 



1. Failure of mass selection and ear-to-row selection beyond the level of 

 the adapted variety. 



2. Crossbreeding recombinations of parent lines of elite hybrids yield 

 little more than the original varieties. 



3. Hybrids of second-cycle and third-cycle lines yield little more than those 

 of the first cycle. 



4. Homozygous corn yields 30 per cent as much as heterozygous corn. 



5. No evidence of epistasis in corn yield. 



6. Regression analyses of yields of Fi's and inbred parents indicate a zone 

 of nearly level regression near the upper end of the range of present data, 

 where it might be predicted with the kind of artificial selection which has 

 been practiced, and in the event of overdominance. 



7. There is some evidence that selection for general combinability alone 

 with respect to yield is effective and this too is consistent with the expectation 

 of overdominance theory. 



8. The fact of hybrid corn is hardly to be explained as other than a result of 

 selection for specific combinability, which in turn is manifestly dependent on 

 heterozygosity of corn yield genes. 



My proposal (Hull, 1945a) that recurrent selection for specific combina- 

 bility be given a trial was made on the assumption that recurrent selection 

 for general combinability or for accumulation of dominant favorable genes 

 had been fairly tried in mass selection and subsequently. The tentative con- 

 clusion was that varieties (and breeds perhaps) were near equilibrium, with 

 mean gene frequencies approximately at (1 + k)/2k. Regression analyses a 

 little later indicated that the corn samples were below equilibrium. Since then 

 it has been proposed orally many times that two parallel breeding plans re- 

 stricted respectively to specific and to general combinability might well be 

 run with corn and with small laboratory animals as pilot experiments. I have 

 later come to believe that recurrent selection among homozygotes might 

 also provide results of considerable theoretical interest. 



